Addison Road

Here’s why I have a huge boycrush on Google -

A crucial watershed for human / machine interface is going to be the development of contextual meaning. Right now, if you say the word “fish”, machines are very good at giving you 12 definitions for the word, 6 million examples of where it was used, the etymology, the brute translation into 35 other languages, etc.

What machines are very bad at doing is understanding the difference between “I was fishing for a compliment” and “I was fishing for trout.” This is contextual meaning, and one promising route for developing this kind of interaction has to do with vast language data mining, collocating contextual markers (when ‘fish’ shows up in verb form near ‘compliment’, it has significance for the meaning) and response analysis (what response did this contextual meaning elicit, and how does that impact our understanding). Google’s underlying technology looks to be very promising at doing this.

This is the brass ring of seamless human / machine agent interaction - getting an interface that has robust contextual recognition, that can learn your own contextual patterns, that can interact with other machine agents to develop meta-data to project potential meanings. Imagine a machine agent that can work through this internal process:

“My owner recently completed a master’s degree in theology at this school, with these other people in his class. He seems to be using the word ‘emerging’ more frequently in his writing and conversations. There’s a good chance that these other people who share his educational context and his social context are using this term in the same way. I should go mine the linguistic data from this very narrow, very specific group of people to see if I can develop a contextual meaning for ‘emerging’ when he uses it in certain ways.”

I walk into the room and say, “Beezle (my computer’s name), is it too late to get a flight to that emerging church conference in San Diego?”

Even if I’ve never used the word “emerging” with Beezle before, he has a very rich, very accurate sense of the meaning, and can draw an inference to the exact conference that I’m referring to based on that meaning.

Google’s in a great position to be able to bring this together. Are they headed that way? Take a look at the type of jobs they’re hiring for at the bottom of the google labs page. The one that caught my eye was “natural language processing.”

We’ll see. I’m excited. What an awesome time to be alive.